During last 7 years (between 12/92 - 8/99) 229 patients were diagnosed to suffer from brain glioma by stereotactic biopsy in the Hospital Na Homolce, Prague. Histological subgroups were determined, using the WHO system. Low-grade gliomas: fibrilary astrocytoma (104 pts), protoplasmatic astrocytoma (5 pts), gemistocytary astrocytoma (6 pts), pleomorphogenic xanthoastrocytoma (1 pt), oligodendroglioma (18 pts), ganglioglioma (4 pts), pilocytic astrocytoma (15 pts), as high-grade gliomas: anaplastic astro- cytoma (23 pts) and glioblastoma (53 pts). It was our intention to find a mathematical model which would be able, based on selected clinical, CT and MRI data, to distinguish between low and high grade gliomas. The statistical model was derived from 214 pts. Ten standard clinical and imaging charac- teristics obtained before the stereobiopsy were selected as predictors of glioma grade and entered into statistical analysis. Binary logistic regression analysis with the stepwise (LR) method performed by SPSS statistical software version 9.0 was used to estimate the probability of glioma grade. Testing the proposed model revealed overall 83.2% agreement between the predicted grade and the histologically verified grade of the glioma. The model was tested in 12 new patients. The percentage of agreement between the model predicted and histologically verified glioma grade in these 12 patients was 83.3%.

        Key words: brain glioma, malignancy